Extending established soil hydraulic property models by non-capillary water: A comprehensive model performance test

<p>Understanding and describing the hydrologic function of soils requires adequate models of soil hydraulic properties. Established models for hydraulic properties implicitly assume that water flow occurs only in completely filled soil pores. This simplification is questionable in cases where soils become dry.  Lab measurements have repeatedly shown that under dry conditions, water retention and hydraulic conductivity are dominated by water in thin films. Today, there are some modelling approaches that take into account this so-called non-capillary water. One of these is the simple Peters-Durner-Iden model system (PDI), which extends any basic model of capillary retention and conductivity by a non-capillary counterpart. In the original form, this requires one additional fitting parameter to characterize the magnitude of non-capillary conductivity. Peters et al. (2021) have recently updated the model system (PDIc) to predict the non-capillary conductivity from the water retention curve without increasing the number of adjustable parameters compared to the established models.</p><p>In this contribution we present a comprehensive model performance test of the established capillary models, the original PDI model, and the new PDIc model. The performance test is based on a data collection of soil hydrological variables measured at 500 undisturbed soil samples. The collection contains soil water retention and conductivity data, determined in the laboratory by the evaporation method, supplemented by dew point method data and measurements of saturated conductivity. For each data set we estimated the soil hydraulic parameters for any combination of the three basic models: van Genuchten with m=1-1/n, van Genuchten with a free parameter m, and Fredlund & Xing and the three considerations of non-capillary water: not considered (no PDI), PDI, and PDIc.</p><p>The results showed that the most flexible basic functions generally yielded the best model fits. For example, the Fredlund & Xing model outperformed the two van Genuchten models. Considering non-capillary water by the PDI model system also clearly increased the model performance. The root mean squared errors (RMSE) for the fits of both the retention and the conductivity curve were clearly reduced in the order from no PDI to PDIc to PDI. Remarkably, the PDIc model generally achieved better fits than the established models although it has exactly the same free parameters.</p><p> </p><p>References<br>Peters, A., T.L. Hohenbrink, S.C. Iden, and W. Durner. A simple model to predict hydraulic conductivity in medium to dry soil from the water retention curve. Geophysical Research Abstracts Vol. 23, EGU21-8717, 2021.</p>